Multi-tier Sentiment Analysis System with Sarcasm Detection: A Big Data Approach

Abstract

Social Media is one of the generating sources of big data and analyzing social big data can provide the valuable information. For analyzing the social big data in an efficient and timely manner, the traditional analytic platform is needed to be scaled up. The powerful technique is necessary to extract the valuable information from social big data. Sentiment Analysis can facilitate valuable information by extracting public opinions. The presence of sarcasm, an interfering factor that can flip the sentiment of the given text, is one of the challenges of Sentiment Analysis. In this paper, Multi-tier Sentiment Analysis system with sarcasm detection on Hadoop (MSASDH) is proposed to extract the opinion from large volumes of tweets. To achieve high-level performance of sentiment classification, MSASDH identifies sarcasm and sentiment-emotion by conducting rule based sarcasm-sentiment detection scheme and learning based sentiment classification with Multi-tier architecture. The large amount of tweets is collected by Apache Flume and it is used for system evaluation. The evaluation results show that detecting sarcasm can enhance the accuracy of Sentiment Analysis. Moreover, the results show that the MSASDH is efficient and scalable by decreasing the processing time when adding more nodes into the cluster.